When an index becomes one trade

An index fund is sold as diversification. When the largest ten holdings are roughly a third of the index and most of them are exposed to the same capital spending cycle, the diversification is partly an illusion. This is arithmetic, not a bearish opinion.

How the concentration happened

Market capitalisation weighting means winners get a larger share automatically. A decade of extraordinary earnings growth at a small number of companies, amplified by passive fund flows that buy in proportion to existing weight, produced a level of concentration not seen since before the war.

The important detail is that these companies earned it. Unlike 1999, the largest weights are highly profitable, cash-generative businesses with real earnings. The bull argument is that this is nothing like the dot-com bubble, and on profitability it is correct.

The dependency that is not obvious from the index

Look at what connects them. One sells the accelerators. Several buy them in enormous quantities for their cloud businesses. Others are customers of those clouds. A chip supplier and a lithography supplier sit upstream of all of it. The revenue of the vendor is the capital expenditure of the customer, and the customer's justification for that spending is future revenue from AI products that are still being proven.

That is a chain, and a chain is only as strong as its assumptions. If AI revenue growth disappoints, the customers reduce capital spending, which is the vendor's revenue, which is the earnings that support a large share of the index.

Circular revenue, and how to think about it

A pattern that has drawn scrutiny: chip vendors investing in AI startups that then spend the investment on chips; cloud providers taking equity in model companies that commit to buying cloud capacity. Some of this is ordinary strategic investment, which has always existed. Some of it inflates reported demand, because the same euro appears as an investment on one statement and revenue on another.

The honest position is that the scale of these arrangements is unusual and the accounting is legitimate, and that both can be true. The question to ask is what share of a vendor's growth comes from customers it has funded. That disclosure is thin.

What would actually break it

Not a bad quarter. The things that would matter are structural: a sustained fall in the growth rate of AI capital spending; a demonstration that inference can run adequately on much cheaper hardware; or an interruption in supply from the concentrated manufacturing base discussed in our read on the chip supply chain.

The reflexive part is worth naming. These companies fund much of their spending from operating cash flow, which is high because the market is strong. A downturn in one feeds the other.

What to do with this, practically

Two things, neither of them a prediction. First, know your actual exposure: if you hold a global index fund, a technology fund and shares in a chip company, you own the same trade three times. Second, if you want the sector exposure without the concentration, equal-weighted indices and ex-technology allocations exist and have had a poor decade, which is exactly why they are cheap.

Our hedge mix calculator makes the correlation point visible: adding assets that move together adds less diversification than the number of line items suggests.

What to watch

Watch capital spending guidance from the largest cloud operators, which is the clearest leading indicator for the whole chain. Watch depreciation schedules, because lengthening the assumed life of accelerators flatters earnings and tells you something about expectations. And watch the top-ten weight, which is published and which most investors have never looked up.